Generative AI Course in Ranchi

1.1 The AI Landscape

• What AI is – and what it is not: AI vs.
  automation vs. analytics
• Narrow AI, General AI and Super AI, with real examples
• AI, Machine Learning and Deep Learning: how the three relate

1.2 Core AI Capabilities

• NLP, computer vision, speech and multimodal AI
• Predictive analytics: forecasting and
  recommendation systems
• Where AI already sits in everyday business
  systems

1.3 AI Across Industries

• HR, operations and supply chain, finance and retail
• Healthcare, manufacturing and education

1.4 Building an AI Mindset

• How to spot a genuine AI use case – and a bad one
• Feasibility: is the data available, repeatable
  and reliable?
• Where AI predictably fails, and why humans
  stay in the loop

2.1 How Generative AI Works

• Predictive AI vs. Generative AI; LLMs without the maths
• Tokens, context windows and why they affect cost
• Training vs. fine-tuning vs. inference; temperature and top-p

2.2 The Current Model Landscape

• GPT, Claude, Gemini and Llama: strengths and limitations
• Reasoning models vs. fast models; open vs. closed weights
• Choosing a model by task, cost, latency and
  data sensitivity

2.3 Prompt Engineering Fundamentals

• Anatomy of a prompt: Role, Task, Context, Format, Constraints
• Generation, summarisation, rewriting and extraction
• Structured output: JSON, tables, lists and schemas

2.4 Prompt Frameworks

• RTF (Role-Task-Format) for quick everyday prompts
• CO-STAR (Context, Objective, Style, Tone, Audience, Response)
• CRISPE and RACE for repeatable professional prompts
• TAG (Task-Action-Goal) for process and workflow prompts
• Building a reusable prompt template library for your role

2.5 Core Prompting Techniques

• Zero-shot and few-shot prompting
• Chain of Thought and self-consistency for harder problems
• Prompt chaining, negative prompting and persona control

2.6 Meta Prompting

• Using AI to write, critique and improve your prompts
• Optimisation loops: draft, test, diagnose, refine
• Generating prompt templates and system prompts automatically
• Comparing prompt variants and evaluating what performs better
• Scaling prompts across a team without quality drift

2.7 Limitations & Quality

• Hallucinations and bias: why they happen and how to spot them
• Verifying facts and grounding answers in sources
• When to use AI and when to use human judgement

3.1 Choosing the Right Model

• LLMs vs. Small Language Models; edge and on-device use
• Reasoning vs. fast models: cost, latency and quality trade-offs
• Use-case mapping: which model for which business task

3.2 Embeddings & Semantic Search

• What embeddings are; representing text as vectors
• Vector databases and why similarity beats keyword search
• Chunking strategies and how chunk size changes answer quality

3.3 RAG in Practice

• RAG end to end: ingest, chunk, embed, retrieve, generate
• Building a knowledge base from documents, wikis and tickets
• Improving retrieval: metadata filters, hybrid search, re-ranking
• Grounding answers with citations; common failure modes

3.4 Customisation & Tool Use

• Prompting vs. RAG vs. fine-tuning: choosing the right lever
• Function calling and structured JSON output
• Model Context Protocol (MCP): connecting models to tools and data

4.1 What Agentic AI is

• Autonomy, goal-directedness and adaptability
• Chatbots vs. workflows vs. agents – a clear comparison
• Real examples: research, support and coding agents

4.2 Anatomy of an Agent

• Memory, planning, tool use and reflection
• Levels of autonomy and where to set the limit

4.3 Agent Design Patterns

• ReAct and plan-and-execute for multi-step tasks
• Router and multi-agent supervisor patterns
• Human-in-the-loop checkpoints and approval gates

4.4 Frameworks, Guardrails & Reliability

• LangChain and LangGraph, CrewAI, AutoGen, Agent SDKs
• Error handling, retries and fallback strategies’
• Guardrails, cost control and monitoring agent performance

5.1 Zapier

• Platform overview and interface
• Building your first Zap; multi-step workflows
• AI-powered steps, filters, formatters and utilities
• Hands-on: automate an email-to-task
  workflow

5.2 Make.com

• Visual workflow builder: modules, routes and scenarios
• Advanced routing and error handling
• Data stores, aggregators, scheduling and webhooks
• Hands-on: build a content aggregation workflow

5.3 LangFlow

• What LangFlow is: a visual builder for LLM apps and AI agents
• Interface tour: canvas, components, flows and the playground
• Core components: model, prompt, memory, chain, agent, tool nodes
• Connecting OpenAI, Claude and Gemini; managing API keys
• Data components: file loaders, splitters, embeddings, vector stores
• Building a RAG chatbot flow step by step
• Turning a flow into an agent with tools and a system prompt
• Custom components and Python code nodes for advanced logic
• Testing, debugging and tracing a flow’s execution
• Publishing a flow as an API endpoint and embedding it in an app
• Environment variables, credentials, versioning and sharing
• Hands-on: build and deploy a document
  Q&A assistant

5.4 Lovable & Vibe Coding

• What vibe coding is: building software by describing it
• Lovable overview: projects, chat interface and live preview
• Prompting for apps: describing screens, data and behaviour
• Building your first app: pages, components and navigation
• Styling and branding: themes, layout and responsive design
• Adding a backend with Supabase: tables, records and queries
• User accounts and authentication
• Iterating safely: chat edits, visual edits, undo and version history
• Connecting APIs and third-party integrations
• Publishing: preview links, custom domains and hosting
• GitHub sync and exporting your code
• Knowing the limits: when to hand off to a developer
• Hands-on: build and publish a working internal business tool

5.5 Notion AI

• AI writing, editing and database automation
• Q&A over workspace knowledge
• Hands-on: build an AI-powered internal knowledge base

5.6 Canva AI

• Text-to-design generation and AI brand kits
• Magic Studio tools: resize, erase, edit, content suggestions
• Hands-on: create a branded presentation with AI

5.7 Choosing your stack

• Matching the tool to the task: automation vs. agent vs. app
• Comparing cost, learning curve and ceiling
  • Security, data residency and what not to paste
  into a tool
  • Avoiding vendor lock-in

5.8 Integrating AI into business Processes

• Identifying automation opportunities in your own role
• Workflow mapping and optimisation
• Change management and team adoption
• Measuring ROI of automation

6.1 n8n Fundamentals

• Cloud vs. self-hosted n8n: which to choose and why
• The canvas, nodes, connections, and executions
• How data flows between nodes; reading the JSON view

6.2 Core Nodes & Logic

• Trigger nodes: schedule, webhook, chat and app events
• HTTP Request node for connecting to any API
• IF, Switch, Merge, Set, Loop and Wait nodes
• Expressions and variables for dynamic workflows

6.3 Credentials & Integrations

• Connecting Google Workspace, Slack, Sheets, Notion, CRMs
• Managing API keys and credentials securely
• Working with webhooks from external systems

6.4 AI Nodes in n8n

• The AI Agent node vs. a simple LLM call
• Chat model nodes: OpenAI, Claude, Gemini and local models
• Memory nodes for conversational context
• Tool nodes: search, calculator, HTTP and workflow tools
• Vector store and embedding nodes

6.5 Building your first AI Agent

• Defining the agent’s goal, scope and tools
• Writing the system prompt and setting guardrails
• Adding memory and testing in the chat interface
• Iterating on failures and edge cases

6.6 RAG Inside n8n

• Ingesting documents from Drive, SharePoint or upload
• Chunking, embedding and storing in a vector database
• Wiring retrieval in as a tool the agent can call
• Keeping the knowledge base fresh on a schedule

6.7 Multi-Agent & Sub-Workflows

• Splitting complex jobs across sub-workflows
• Routing between specialist agents
• Adding human approval steps before an agent acts
• Hands-on: a support agent answering from a company knowledge base
• Hands-on: a lead-qualification agent that writes to a CRM
• Hands-on: a daily research digest agent that emails a brief

7.1 The Copilot Landscape

• Microsoft 365 Copilot vs. Copilot Chat vs. GitHub Copilot
• Licensing basics and what each tier includes
• Where your data goes: the tenant boundary explained simply

7.2 How Copilot Works

• Grounding in Microsoft Graph: files, mail, chats, and meetings
• How permissions decide what Copilot can see
• Why Copilot answers differ from public
  ChatGPT answers

7.3 Copilot in Word

• Drafting from a prompt, a file or a set of notes
• Rewriting for tone, length and audience
• Summarising long documents and generating tables

7.4 Copilot in Excel

• Generating and explaining formulas
• Analysing a table: trends, outliers and breakdowns
• Creating charts and PivotTables by description
• Data cleaning and column transformation prompts

7.5 Copilot in PowerPoint

• Generating a deck from a Word document or a prompt
• Restyling and reorganising existing slides
• Generating speaker notes and summarising a deck

7.6 Copilot in Outlook

• Summarising long threads and extracting decisions
• Drafting and refining replies; coaching on tone
• Meeting preparation briefs

7.7 Copilot in Teams

• Meeting recaps, action items and decision logs
• Catching up on missed chats and channels
• Using Copilot during a live meeting

7.8 Prompting Copilot Effectively

• Referencing files, people and meetings inside a prompt
• Prompt patterns that work in a Microsoft 365 context
• Using the Prompt Gallery and building a team prompt library

7.9 Governance & Adoption

• Sensitivity labels, data protection and admin controls
• Common oversharing risks and how to avoid them
• Driving adoption: use cases by department
• Hands-on: build a prompt library and adoption plan for your team

8.1 The Claude Model Family

• Opus, Sonnet and Haiku: capability, speed and cost
• Extended thinking and when deeper reasoning pays off
• Choosing the right Claude model for a given task

8.2 Claude Interface Essentials

• Projects: giving Claude persistent context for ongoing work
• Project knowledge, custom instructions and styles
• Managing long conversations effectively

8.3 Working with Long Context

• Analysing large documents, contracts and reports
• Comparing multiple documents in a single pass
• Synthesis and structured extraction tasks

8.4 Artifcats

• Generating documents, dashboards and mini-apps in chat
• Iterating on an artifact through conversation
• Business uses: trackers, calculators, one-pagers

8.5 Claude – Code Overview

• Agentic coding from the terminal, IDE or desktop
• Repository-aware tasks and multi-file changes
• Where it fits for non-developers: scripts, automation, data tasks

8.6 Connectors & MCP

• Model Context Protocol in practice
• Connecting Claude to Drive, calendar, email and internal tools
• Workflows that read from and write to business systems
• Permissions and safe connector use

8.7 Prompting Claude Well

• Using XML tags to structure complex prompts
• Role framing, examples and explicit output formats
• Encouraging step-by-step reasoning; controlling verbosity
• How Claude prompting differs from GPT prompting

8.8 Claude in Business Workflows

• Research, analysis, drafting and document review
• Building a repeatable assistant for a business function
• Data handling, confidentiality and safe use
  policies
• Hands-on: build a Claude Project with connected data sources

9.1 The Visual Gen AI Landscape

• Image models, video models and avatar platforms
• Quality, speed and cost trade-offs across the field
• Choosing a tool for marketing, training or product work

9.2 AI Image Generation

• Google Nano Banana and Nano Banana Pro (Gemini image models)
• ChatGPT image generation; Midjourney; Stable Diffusion
• Strengths compared: photorealism, text-in image, speed, control
• Resolution, aspect ratios and output formats

9.3 Prompting for Images

• Prompt anatomy: subject, style, composition, lighting, camera
• Negative prompts and constraint control
• Reference images and style transfer
• Getting readable text inside an image

9.4 Editing & Brand Consistency

• Inpainting, outpainting and background replacement
• Multi-reference fusion and pose control
• Keeping a character, product or brand look consistent
• Building a reusable brand prompt kit

9.5 AI Video Generation

• Google Veo: text-to-video and image-to video with native audio
• Gemini Omni Flash for fast conversational video editing
• Kling and Runway as alternative production tools
• Practical limits: clip length, continuity and cost

9.6 Prompting for Video

• Shot description, camera movement and pacing
• Image-to-video: starting from a generated still
• Maintaining continuity across multiple shots
• Storyboarding an AI video before you generate

9.7 AI Avatars & Talking – Head Video

• Avatar platforms: HeyGen and Synthesia
• Creating a custom or stock avatar; multilingual delivery
• AI voice with ElevenLabs: cloning, tone and pacing
• Script to avatar to voice to captions: the full pipeline
• Hands-on: a branded image campaign set with one visual identity
• Hands-on: a 60-second avatar-led explainer with voice and captions

10.1 The Real Risks

• Hallucinations and factual failure
• Bias in training data, algorithms and human feedback
• Privacy, data leakage and prompt injection
• Deepfakes, misinformation and synthetic media

10.2 Responsible AI Principles

• Fairness, accountability, transparency and explainability
• Privacy, safety and human control

10.3 Frameworks at a glance

• OECD AI Principles and NITI Aayog #AIForAll (India)
• UNESCO AI Ethics, the EU AI Act, MicrosoftnResponsible AI

10.4 Safe Usage in Practice

• Verifying outputs and cross-checking sources
• Human-in-the-loop: when oversight is mandatory
• Guardrails: input validation, output moderation, action limits
• Secure handling of company and customer data

10.5 Lessons & Culture

• Short case studies of AI failures and what they teach
• Individual and organisational accountability
• Building responsible AI habits into daily work

Trends to Watch

• Multimodal and on-device AI
• AI coding assistants: GitHub Copilot, Cursor,
  Claude Code
• The MCP and agent-tooling ecosystem
• What the Indian job market is hiring for in AI right now